Architectural Enumeration and Evaluation for Identification of Low-Complexity Systems

The cost of large complex systems can be reduced by using system complexity as a cost proxy during the initial stages of system architecting. This paper presents Architectural Enumeration and Evaluation (AEE), a method for rapid, efficient and thorough consideration of enormous architectural design spaces to find the best low-complexity solutions. AEE assembles promising system architectures from any number of candidate technologies and evaluates them relative to a set of customer-value metrics, which can include a complexity metric as a proxy for cost. This paper also presents an example application of AEE together with a complexity metric and a spectral graph partitioning method, to enumerate the feasible set of architectures, and identify for each architecture its lowest complexity hierarchical clustering of subsystems. I. Introduction ignificant technological and architectural changes have been introduced into aerospace systems, over the past several decades, in an effort to improve the performance and capability of new platforms. These changes have led to an exponential growth in the complexity of modern aerospace platforms and accompanying design and development challenges. These increases in technical complexity have been accompanied by increases in the complexity of system requirements and organizational partnerships. Complexity in requirements stems from the need to meet multiple present and future mission requirements. Analysis by the RAND Corporation found that the largest component in the growth in the cost of fixed-wing aircraft has come from increased complexity. [7] As systems have become more complex their development cost has increased and the predictability of their development cost has decreased. Managing and minimizing complexity of new system development offers the ability to reduce the magnitude and increase the predictability of development cost. Fundamental architectural decisions made early in the design process have a major impact on system complexity; however, system architects lack quantitative feedback on the complexity of their architectures. During the early design stages of large systems, systems are architected in terms of functional subsystems (e.g., Electrical Power System, Environmental Control System, Auxiliary Power Unit) and their interconnections as shown in Figure 1. This standard top-down design approach is a divide-and-conquer strategy that defines the subsystem boundaries as an interface to decouple the overall design problem into two sequential sub-problems. The first sub-problem is to design a system of functional subsystems based on overall system requirements, resulting in a set of requirements for each subsystem. The second sub-problem is to design each functional subsystem in terms of its required functionality. This approach has also enabled airframers to focus on the large-scale system-integration problem, while outsourcing subsystem design to subcontractors, with the subsystem requirements serving as the organizational interface. Since these two sub-problems are not really decoupled, subsystem interactions necessitate design iteration to meet system-level requirements. Because functional subsystems are designed separately, these iterations often occur late in the design process and are a significant cause of schedule and cost overruns. One approach to design an overall system comprised of functional subsystems (the first sub-problem) is to enumerate and evaluate all feasible solutions to this problem, as described in [1]. This approach has typically been avoided due to the exponential size of the design space. However, it has recently been shown that the feasible set is